[
https://issues.apache.org/jira/browse/HBASE-21336?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16658369#comment-16658369
]
stack commented on HBASE-21336:
-------------------------------
bq. I do not think we need to guarantee any order when loading procedures.
bq. But now I sort the procedures so the parent procedure will arrive at first
then the test will always pass and then we will set it to RUNNABLE...
bq. And I think the force update logic can also make this happen, as it will
update the parent procedure if it is stuck there, and mess up the replay
order(for the old implement). Let me see how to fix this.
So, did you drop ordering? You just have a sort on load? What did you do for
force update?
Would be sweet if you could do without ordering.
Looking at patch...
> Simplify the implementation of WALProcedureMap
> ----------------------------------------------
>
> Key: HBASE-21336
> URL: https://issues.apache.org/jira/browse/HBASE-21336
> Project: HBase
> Issue Type: Sub-task
> Components: proc-v2
> Reporter: Duo Zhang
> Assignee: Duo Zhang
> Priority: Major
> Fix For: 3.0.0, 2.2.0
>
> Attachments: HBASE-21336-v1.patch, HBASE-21336-v2.patch,
> HBASE-21336-v3.patch, HBASE-21336.patch
>
>
> I do not think we need to implement the logic from such a low level, i.e,
> building complicated linked list by hand, which makes it really hard to
> understand.
> Let me try to implement it with existing data structures...
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
